Goldman Sachs partner warns: Risks in the AI ecosystem "can no longer be ignored"

AI Investment Boom Is Accumulating Systemic Vulnerabilities Goldman Sachs partner Rich Privorotsky issues a warning: As market positioning, leverage, and AI capital expenditures become deeply intertwined, the inherent cyclical risks within this ecosystem are increasingly unavoidable. If the AI spending cycle is disrupted, the market's fragility will be fully exposed. The immediate catalyst for market alert is that data center developer Crusoe has suspended a 1.8-gigawatt data center project in Wyoming—the halt was at the request of its client, whose identity has not been disclosed. Privorotsky points out that in a market where almost all asset pricing is tied to AI capital spending, even a single project delay or reprioritization is enough to prompt investors to reassess their predictions for future demand. Meanwhile, the Nasdaq index has fallen about 6% since last Thursday’s close, and the performance of stocks has lagged notably behind bonds and crude oil. Privorotsky warns that momentum returns are at the 90th percentile over the past five years, overall market exposure is at the 99th percentile, leverage demand is widening financing spreads, and retail participation through leveraged ETFs remains substantial—these combined factors make the market highly sensitive to any disruption in the AI spending cycle. Data Center Project Suspension Puts AI Demand Assumptions Under Scrutiny Crusoe’s suspension of the Wyoming project attracts attention because of its unique design—using 900 megawatts of behind-the-meter energy to bypass power grid bottlenecks, specifically addressing power supply challenges for AI computing infrastructure. The project was halted during its initial development phase, and the client's identity remains undisclosed. Privorotsky states clearly that drawing macro conclusions from this event would be a mistake. Nevertheless, the symbolic significance of this event should not be overlooked. In current market narratives, AI hardware and data center investments are the core pillars supporting tech stock valuations. Any signal of slowing demand, project delays, or shifting client priorities will be amplified by the market. Privorotsky’s assessment: Isolated cases do not constitute a trend, but in an overcrowded market structure, tolerance for such signals has greatly diminished. AI Ecosystem Divergence: Parallel Tracks of Frontier Intelligence and Local Models Privorotsky describes a "dual-mode world" that is taking shape: one end has centralized, cloud-accessed, expensive frontier models and possible future superintelligence; the other has virtually free, open, and increasingly powerful local AI layers that handle most everyday tasks. He predicts the ultimate form will be basic work done on local or open-source models, while complex reasoning and problem solving are handled by high-end cloud systems. There are fundamental disagreements over what this divergence means for the AI investment cycle. The optimistic interpretation is that the overall pie will keep growing, with rising demand for edge computing, data centers, storage, power, and networking; pessimists believe most economically valuable tasks could soon or already be run on existing hardware, meaning actual demand turning points may lag far behind timelines implied by current valuations. Privorotsky notes that the focus of debate is increasingly less about model quality itself, and more about where inference computation ultimately occurs. Cyclical Risk Accumulation, Leverage Exposure Amplifies Vulnerabilities Privorotsky’s concerns about the current market structure focus on a central judgment: AI spending has become the market’s "long leg"—it is the main driver of hardware investment, a significant contributor to GDP growth, and a key pillar supporting overall market performance. This highly concentrated dependency forms a self-reinforcing cycle, whose fragility is becoming ever more obvious. He especially notes that the large embedded and synthetic exposures created by leveraged products have led to a severe undervaluation of market hedging tools. In his view, holding gamma (convexity exposure in options) holds much more practical value in current portfolio management than most investors realize. This risk is not new, but the market is accelerating towards it rather than actively avoiding it. AI Investment-Driven Economy Is "Overheating" On the macro level, Privorotsky turns his attention to U.S. CPI data. The U.S. Bureau of Labor Statistics reported on Wednesday that the Consumer Price Index (CPI) rose 4.2% year-on-year in May, the highest since early 2023 and in line with market expectations. Privorotsky stresses that the relatively mild official data does not hide the strong inflationary pressure in reality: rising energy costs, nearly one trillion dollars in AI-related business investment, and a U.S. fiscal deficit accounting for about 6%–7% of GDP together create structural drivers for inflation. His view is: this is an investment-driven, overheating economy. Unless AI brings sufficiently significant productivity gains, inflation will persist as a long-term byproduct. Risk Warning & Disclaimer The market carries risks; investment should be approached with caution. This article does not constitute personal investment advice and does not take into account individual users’ special investment targets, financial status, or needs. Users should consider whether any opinions, views, or conclusions provided here suit their specific circumstances. Investments made accordingly are at one’s own risk.